Information-theoretic approach to high-temperature spin dynamics

Abstract

The information-theoretic maximum-entropy approach is used to obtain high-temperature dynamical spin-correlation functions from a finite number of rigorously known moments. The method is applied to the dipolar-broadened magnetic-resonance line-shape function for a simple cubic lattice and other spin-correlation functions. The results improve on the best previous theoretical results and agree with experiment to within ∼2%. Modeling of the self-energy with the maximum-entropy method provides better results than direct modeling of the line-shape function. The results display an oscillating pattern of convergence which is expected to occur in many physical applications.